El caso particular de los c´ odigos convolucionales de Goppa de

This study, which was conducted at Uludağ University, School of Foreign Languages, examined to the extent to which listening to audiobook affects students’ pronunciation skills on both recognition and production levels and how students perceive the use of audiobooks with the aim of teaching and learning pronunciation in English.

The research questions addressed in the study were as follows: 1. What are the perceptions of EFL students about using audiobooks to improve their pronunciation?

2. What are the effects of listening to audiobooks on EFL students’ a. recognition level of problematic phonemes in English?

b. pronunciation of problematic phonemes in English?

3. Does the effect of listening audiobooks on pronunciation differ in terms of the proficiency levels of students?

Data Analysis Procedures

After the data collection process was completed with pre- and post-treatment

recognition tests, pre- and post-treatment production tests and pre- and port-treatment questionnaires, the first step in the data analysis was to score the participants’ pre and post-treatment production tests’ scores. Once all the scores of 65 participants were obtained, the data of pre and post-treatment production tests’ scores were entered into Statistical Package for Social Sciences (SPSS). Since most of the data was automaticallytransferred into SPSS from Limesurvey.com, the main data collection tool for the study, manual data entering process was not implemented for the two

other datasets (pre- and post-treatment recognition tests and pre- and post-treatment questionnaires). The participants from three different proficiency levels were put together in the aforementioned three datasets, but labeled differently in accordance with the level (1= Elementary, 2= Pre-intermediate, 3= Intermediate). After these adjustments were made, a Paired samples t-test analysis was conducted to compare the pre- and post-treatment questionnaires. Later, a Paired samples t-test analysis was run to see the difference between the pre- and post-test results for the three different proficiency groups in recognition and production tests. Lastly, One-way Analysis of Variance (One-way ANOVA) was run to see if the participants’ level of scores on both recognition and production levels changed depending upon their language proficiency level.

Results

The results will be presented in accordance with the research questions of the study. First, the answer to the research question 1, “What are the perceptions of EFL students about using audiobooks?” will be revealed, then the answer to the research question 2a, “What are the effects of listening to audiobooks on EFL students’ recognition level of problematic phonemes in English?” will be introduced, next, the answer to the research question 2b, “What are the effects of listening to audiobooks on EFL students’ pronunciation of problematic phonemes in English?” will be presented, lastly, the answer to the research question 3, “Does the effect of listening audiobooks on pronunciation differ in terms of the proficiency levels of students?” will be discussed.

What are The Perceptions of EFL Students about Using Audiobooks to Improve Their Pronunciation?

In order to examine the difference between the participants’ pre and post- treatment perceptions of the effects of listening to audiobooks on their pronunciation,

first, the descriptive statistics of the pre and post-questionnaire scores were calculated. Since two outliers were detected and the inspection of their values

revealed them to be extreme, they were excluded in line with the recommendation by Bakker and Wicherts (2014). A Shapiro-Wilk's test (p > .05) and a visual inspection of their histograms, normal Q-Q plots and box plots showed that the test scores were approximately normally distributed for each level of proficiency, with a skewness of -0.64 (SE= 0.49) and a kurtosis of -0.39 (SE= 0.95) for elementary level; a skewness of 0.24 (SE= 0.50) and a kurtosis of -1.31 (SE= 0.97) for pre-intermediate level, and a skewness of -0.61 (SE= 0.51) and kurtosis of 0.13 (SE= 0.99) for intermediate level. The responses to the twenty- three items of the questionnaire about the attitudes towards audiobooks and their effects on pronunciation were scored as 1= Strongly disagree, 2= Disagree, 3= Undecided, 4= Agree, 5= Strongly agree. That is to say a higher mean represents a more positive attitude of the participants. Figure 1 shows the means of participants’ pre and post-treatment scores for the perception questionnaire.

Looking at this figure and the descriptive statistics, none of the proficiency levels has a completely positive attitude towards the effects of audiobooks on

pronunciation. In a similar way, none of the proficiency levels has a negative attitude towards the effects of audiobooks on pronunciation. However, the post-treatment questionnaire mean of pre-intermediate and intermediate levels were slightly higher than their pre-treatment questionnaire means, whereas the pre and post-treatment questionnaire means of the elementary level remain the same (𝑥 ̅= 4.02). While the mean of the pre-treatment questionnaire was 3.89 and the mean of the second production test was 3.95 for pre-intermediate level, the mean of the first production test was 3.71 and the mean of the second production test was 3.76 for intermediate level. The descriptive statistics also revealed that the highest positive attitude level towards the effects of audiobooks on pronunciation was at elementary level, whereas the least positive attitude level towards the effects of audiobooks on pronunciation was at intermediate level. As a result of this, a parametric paired-sample t-test was conducted to evaluate whether the differences existed between the means of the pre and post-treatment questionnaire scores were statistically significant (see Table 4). Table 4

Difference between the Pre and Post-treatment Attitude Questionnaire Items for All Levels

p > .05 level.

The results of the paired sample t test revealed that the mean difference between the pre and post-treatment means for these items on the questionnaire were not statistically significant. Participants’ pre and post- treatment perspectives towards the effects of audiobooks on pronunciation differed very little from one another with

Scores 𝑥 ̅ SD df t p

Pre 3.92 0.32 62 0.55 0.587

a mean of 0.02, t (62) = 0.55, p > .05, d = 0.06, indicating that there is not a statistically significant difference in attitudes prior to and after the treatment. (𝑥 ̅= 3.92, SD= 0.32) and (𝑥 ̅= 3.95, SD= 0.35). The effect size (d = 0.06) was small based on Cohen’s conventions (1988).

What are the Effects of Listening to Audiobooks on EFL Students’ Recognition Level of Problematic Phonemes in English?

In order to examine the difference between the participants’ first and second recognition test scores, first, the descriptive statistics were calculated. A Shapiro- Wilk's test (p > 0.05) and a visual inspection of their histograms, normal Q-Q plots and box plots revealed that the normality assumption was not violated, with a

skewness of 0.34 (SE= 0.50) and a kurtosis of -0.34 (SE= 1.00) for elementary level; a skewness of -0.18 (SE= 0.50) and a kurtosis of 0.18 (SE= 1.00) for pre-

intermediate level, and a skewness of 0.50 (SE= 0.50) and kurtosis of -0.70 (SE= 1.00) for intermediate level. Figure 2 shows the means of participants’ first and second test scores of the recognition test.

According to the descriptive statistics, the second recognition test score means of all the proficiency levels were higher than their first recognition test score means. While the mean of the first recognition test scores was 0.56 and the mean of the second recognition test scores was 0.57 for elementary level, the mean of the first recognition test scores was 0.54 and the mean of the second recognition test scores was 0.65 for pre-intermediate level, and the mean of the first recognition test scores was 0.56 and the mean of the second recognition test scores was 0.65 for

intermediate level. As a result of this, a parametric paired- sample t test was conducted to evaluate whether this increase between the means of the recognition test scores before and after listening to audiobooks was statistically significant (see Table 5).

Table 5

The Mean Difference between the First and Second Recognition Test of All Levels

p < .001 level.

The results of the paired sample t test elicited a statistically significant mean increase of 0.07, t (64) = 6.82, p < .001, d = 0.84, indicating that there was a

statistically significant increase in the recognition test scores from the first to the second (𝑥 ̅= 0.55 to 𝑥 ̅= 0.62). The effect size (d = 0.84) was small based on Cohen’s conventions (1988). This means that the participants’ performance of recognizing problematic phonemes improved overall after listening to the audiobook.

A second set of tests was run in order to investigate the differences that occurred between the first and second recognition tests for each problematic

phoneme. First, words were grouped in accordance with the phoneme they represent Scores 𝑥 ̅ SD df T p

1st Test .55 .08 64 6.82 .000

(ð, ə, n, ŋ, ɛ, θ). Second, the means of the phonemes tested before and after

audiobook listening session were calculated. Then a parametric test, t-test was used to determine the differences in the first and second performances. The results of the paired sample t-test that was run to see the overall differences that occurred between the first and second recognition test scores for each phoneme is presented in Table 6. Figure 3 presents the first recognition test means of each phoneme for each

proficiency level, and Figure 4 presents the second recognition test means of each phoneme for each proficiency level.

Table 6

The Mean Difference of All Levels between the First and Second Recognition of The Phonemes Phonemes _{𝑥 ̅ SD t df p} ð _{-.02 .21 -0.82 64 .410} ə _{-.02 .22 -0.92} 64 _.359 θ _{-.05 .26 -1.55} 64 _.126 ɛ _{-.10 .26 -3.00} 64 _.004* ŋ _{-.06 .22 -2.31} 64 _.024** n _{-.14 .24 -4.89} 64 _.000* p < .001* p > .05**

The results of the paired sample t test indicated that there was a mean increase in the differentiation of all phonemes after the audiobook listening session. However, the results indicated that this increase was statistically significant for only three phonemes /ɛ/ (𝑥 ̅= -0.10, SD= 0.26), /ŋ/ (𝑥 ̅= -0.06, SD= 0.22), and /n/ (𝑥 ̅= - 0.14, SD= 0.24). The results of the t test suggest that the participants’ performance of recognizing problematic phonemes of /ɛ/, /ŋ/ and /n/ has improved to the statistically significant level after the audiobook treatment.

Figure 3. Mean range of the phonemes in first recognition test.

Examination of the first recognition test histograms revealed that for

elementary level, the lowest performance was for the recognition of the phoneme /ð/ (𝑥 ̅ = 0.48), while the highest performance was at for recognition of the phoneme /θ/ (𝑥 ̅ = 0.61). At pre-intermediate level, the lowest performance was observed for the recognition of the phoneme /ɛ/ (𝑥 ̅ = 0.47), whereas the participants showed the highest performance for the recognition of the phoneme /ə/ (𝑥 ̅ = 0.61). Intermediate level participants’ performance was lowest in the recognition of the phonemes /ð/ and /θ/ (𝑥 ̅ = 0.51), while they were able to identify the phoneme of /ə/ at the highest level (𝑥 ̅= 0.63).

Figure 4. Mean range of the phonemes in second recognition test.

The investigation of the second recognition test histograms revealed that for the elementary level, the lowest performance was the recognition of the phoneme /ɛ/ (𝑥 ̅ = 0.52), while the highest performance was for the phoneme /n/ (𝑥 ̅ = 0.68). At pre-intermediate level, the lowest performance was for the recognition of the phoneme /ð/ (𝑥 ̅ = 0.54), whereas the participants scored highest for the recognition of the phoneme /n/ (𝑥 ̅ = 0.75). Intermediate level participants’ performance was lowest for the phonemes /ð/ (𝑥 ̅= 0.53), while they were able to identify the phoneme of /ɛ/ at the highest level (𝑥 ̅ = 0.78).

What are The Effects of Listening to Audiobooks on EFL Students’ Pronunciation of Problematic Phonemes in English?

In order to examine the difference between the participants’ first and second production test scores, first, the descriptive statistics were calculated. A Shapiro- Wilk's test (p > 0.05) and a visual inspection of their histograms, normal Q-Q plots

and box plots showed that the test scores were approximately normally distributed for each level of proficiency, with a skewness of -0.20 (SE= 0.50) and a kurtosis of - 0.54 (SE= 1.00) for elementary level; a skewness of 0.13 (SE= 0.49) and a kurtosis of -0.84 (SE= 1.00) for pre-intermediate level, and a skewness of -0.17 (SE= 0.51) and kurtosis of -0.61 (SE= 1.00) for intermediate level. Figure 5 shows the means of participants’ first and second test scores of the production test.

Figure 5. First and second production test means of the proficiency levels

According to the descriptive statistics, the second production test means of all the proficiency levels were higher than their first production test means. For the elementary level, the mean of the first production test was 0.50 and the mean of the second production test was 0.64. The mean of the first production test was 0.60 and the mean of the second production test was 0.74 for pre-intermediate level. For the intermediate level, there was the greatest difference, with the mean of the first production test was 0.56 and the mean of the second production test was 0.73 for

intermediate level. As a result of this, a parametric paired- sample t-test was

conducted to evaluate whether this increase was statistically significant (see Table 7). Table 7

The Mean Difference between the First and Second Production Test of All Levels

p < .001

The results of the paired sample t-test elicited a statistically significant mean increase of 0.15, t (64) = 15.31, p < .001, d = 1.90, indicating that there is a

statistically significant mean increase in production test scores from the first production test (𝑥 ̅= 0.55, SD= 0.09) to the second production test (𝑥 ̅= 0.70, SD= 0.11). The effect size (d = 1.90) was small based on Cohen’s conventions (1988). Put another way, the results suggest that the participants’ performance of producing problematic phonemes has improved significantly.

Another series of test was conducted to investigate the differences that occurred in the production of each problematic phoneme before and after the

treatment. The means for both the pre-treatment and post-treatment production tests for each phoneme (ð, ə, n, ŋ, ɛ, θ) were calculated. Then t test was used to determine the differences in the first and second performances. The results of the paired sample t test that was run to see the overall differences occurred between the first and second production test scores for each phoneme is presented in Table 8. Figure 6 presents the first production test means of each phoneme for each proficiency level, and Figure 7 presents the second production test means of each phoneme for each proficiency level.

Scores 𝑥 ̅ SD df T p

1st Test .55 .09 64 15.31 .000

Table 8

The Mean Difference of All Levels between the First and Second Recognition of the Phonemes Phonemes 𝑥 ̅ SD T df p ð -.22 .20 -9.00 64 .000* ə -.81 .22 -29.00 64 .000* θ -.07 .17 -3.33 64 .001** ɛ .06 .29 1.63 64 .108 ŋ .25 .22 9.09 64 .000* n -.20 .30 -6.00 64 .000* p < .001* p < .01**

The results of the paired sample t-test show that there is a mean decrease in the differentiation of the phonemes /ɛ/ and / ŋ/, while there is a mean increase for the phonemes of /ð/, /ə/, /n/ and /θ/ after the audiobook listening session. The results indicated that these differences are statistically significant for five of the phonemes /ð/ (𝑥 ̅= -.22, SD= .20), /ə/ (𝑥 ̅= -.81, SD= .22), /n/ (𝑥 ̅= -.07, SD= .17), /ŋ/ (𝑥 ̅= .25, SD= .22) and /θ/ (𝑥 ̅= -.20, SD= .29). That is to say, the results of the t test suggest participants’ performance in producing four of the problematic phonemes /ð/, /ə/, /n/ and /θ/ improved for the phonemes and decreased for the phoneme /ŋ/ at a

Figure 6. Mean range of the phonemes in first production test

The investigation of the first production test histograms revealed that for elementary level, the lowest performance was for the production of the phoneme /θ/ (𝑥 ̅= 0.12), while the highest performance was for the phoneme /ŋ/ (𝑥 ̅ = 0.84). At pre-intermediate level, the lowest performance was observed for the recognition of the phoneme /θ/ (𝑥 ̅ = 0.19), whereas the participants scored highest performance on the phoneme /ŋ/ (𝑥 ̅ = 0.92). Intermediate level participants’ performance was lowest in the recognition of the phoneme /ə/ (𝑥 ̅ = 0.11), and highest for the phoneme of /ŋ/ (𝑥 ̅ = 0.90).

Figure 7. Mean range of the phonemes in second production test.

An examination of the second production test histograms revealed that for all three levels (elementary, pre-intermediate, and intermediate) level, the lowest

performance was for the phoneme /θ/ (𝑥 ̅ = 0.28, 𝑥 ̅= 0.40, 𝑥 ̅= 0.41, respectively). For all three levels, the sound produced at the highest level was also

the same, /n/, with means of 0.98, 0.40, and 0.95 for the elementary level, pre-intermediate, and intermediate levels respectively.

Does The Effect of Listening Audiobooks on Pronunciation Differ in Terms of The Proficiency Levels of Students?

In order to examine whether there is a difference among the proficiency levels in terms of their performance in pronunciation recognition test, first, the gain score was calculated by subtracting the mean of the first recognition test score from the mean of the second recognition test score. Then, the descriptive statistics for the pronunciation recognition gain score were calculated. After one outlier from the

elementary level was excluded in line with the recommendation by Bakker and Wicherts (2014), tests revealed that pronunciation recognition test scores were normally distributed, as assessed by Shapiro-Wilk's test (p > .05). Since the normality assumption could be met, a One-Way Analysis of Variance (One-Way ANOVA) was used to determine the relationships of each of three variables with participants’ pronunciation recognition scores. The results of the descriptive statistics and the mean differences are given in Table 9 and Figure 8.

Table 9

Results of the Descriptive Statistics on Pronunciation Recognition Test Scores

Group N 𝑥 ̅ SD

Elementary 21 .02 .03

Pre-intermediate 22 .11 .09

Intermediate 21 .09 .06

The assumption of homogeneity of variances was violated, as assessed by Levene's test for equality of variances (p = .0005). As a result of this, Welch ANOVA, which adjusts for the lack of homogeneity, was used to explore the differences and it is revealed that the ability to recognize problematic phonemes (differences between the pronunciation recognition pre- and post-test scores) was statistically significant for the proficiency levels, Welch's F(2, 37.10) = 16.09, p < .0005. Post-hoc tests were also conducted to investigate where any difference occurred (See Table 10).

Table 10

Comparison between Levels for Differences of Score Means on Pre- and Post- Recognition Test

Proficiency Levels 𝑥 ̅ SE p

Elementary - Pre –intermediate -.09 .02 .001*

Elementary – Intermediate -.07 .01 .000**

Pre-intermediate – Intermediate .01 .02 .709 p < .01*

p < .001**

The results in Table 9 compare the differences on the pre- and post-treatment recognition tests between groups. That is, for the first line, 𝑥 ̅indicates the difference between the mean gains between the elementary level (.02 as reported in Table 8) and the pre-intermediate level (.11 in Table 8). The results (𝑥 ̅ = -0.09, SE = 0.02), indicate that the difference in mean gains is negative (i.e., the pre-intermediate level had a larger gain) and that this difference is statistically significant (p < .01). Similar results were found in comparing the mean gains for the intermediate level to the elementary level (𝑥 ̅ = -0.07, SE = 0.01). Again, the sign indicates that the gain was greater for the intermediate level, while the difference between the groups’ gains was also found to be significant (p < .001). A smaller difference was calculated for the

differences in means between the pre-intermediate level to intermediate level (𝑥 ̅ = 0.01, SE = 0.02) and this difference was not significant (p >.05).

In order to examine whether there is a difference among the proficiency levels in terms of their performance in pronunciation production test, first, the gain score was calculated by subtracting the mean of the first production test score from the mean of the second production test score. Then, the descriptive statistics of the pronunciation production gain score were calculated, and it was revealed that pronunciation production test scores were normally distributed, as assessed by Shapiro-Wilk's test (p > .05). Since the normality assumption was met, a One-Way Analysis of Variance (One-Way ANOVA) was used to determine the relationships of each of three variables with participants’ pronunciation production scores. The results of the descriptive statistics and the mean differences are given in Table 11 and Figure 9.

Table 11

Results of the Descriptive Statistics on Pronunciation Production Test Scores

Group N 𝑥 ̅ SD

Elementary 22 .14 .07

Pre-intermediate 22 .14 .06

Figure 9. Production test gain score mean differences of the proficiency levels

The assumption of homogeneity of variances was violated, as assessed by Levene's test for equality of variances (p = .0005). As a result of this, Welch

ANOVA was again used to examine the differences. The results of this test revealed that the ability to produce problematic phonemes (pronunciation recognition test score) was not statistically significantly different for different proficiency levels, Welch's F(2, 39.410) =.973, p = .387. Post-hoc tests were also conducted to investigate where any difference occurred. The Table 12 shows the results of post- hoc tests.

Table 12

Comparison between Levels for Differences of Score Means on Pre- and Post- Production Test

Proficiency Levels 𝑥 ̅ SE p

Elementary- Pre –intermediate .00 .02 0.974

Elementary – Intermediate -.03 .03 0.484

Pre-intermediate – Intermediate -.04 .03 0.360 p > .05

The results in Table 11 compare the differences on the pre- and post-

treatment production tests between groups. That is, for the first line, 𝑥 ̅indicates the difference between the mean gains between the elementary level (.14 as reported in Table 10) and the pre-intermediate level (.14 in Table 10). The results (𝑥 ̅ =

0.00, SE = 0.02), indicate that there is a slight difference in mean gains (i.e., both the